📩 Aperture@thegaogroup.com
Due to our team’s recent breakthroughs in physical AI and AIoT, this website and the websites of all 64 portfolio companies are being substantially upgraded to incorporate our Physical AI and AIoT engines and sensing technologies. This overhaul will be completed by Nov. 30, 2026.
This page has not been updated yet.
Equipment Health Inspection and Quality Intelligence
Develop diagnostics and inspection methods that help operators detect deterioration, explain defects and prioritize intervention. This program includes rotating equipment, manufactured parts and civil infrastructure.
This research program is conducted by a team of Ph.D. researchers with expertise in machine inspection and quality intelligence.
Transferable Predictive Maintenance and Equipment Diagnostics
Combine vibration, temperature, acoustic, current and operating data to identify degradation across changing loads. Investigate self supervised learning, domain adaptation and physics informed features where failure labels are scarce. Estimate remaining useful life only where degradation histories and validation conditions support it.
Research questions
How much new data is needed for another machine? Can the model separate workload changes from deterioration and abstain when unfamiliar behavior exceeds its experience?
Proposed Demonstration
Train on a motor, pump or fan rig and test on another unit and unseen operating regimes; hold out entire machines from training.
Evaluation
Missed faults, false alarms per operating hour, detection lead time, diagnostic accuracy, adaptation effort and prediction interval coverage.
Multimodal Industrial Inspection and Defect Detection
Combine visible images with thermal, acoustic, vibration or depth observations. Study small defects, unusual surface finishes, changing illumination and incomplete inspection coverage. Evaluate novelty detection and image suitability checks before adding robot motion to an inspection process.
Research questions
When does another modality improve defect detection enough to justify its cost? Can a model recognize an unfamiliar defect without treating every harmless variation as a failure?
Proposed demonstration
Inspect a defined assembly at a fixed station with varied lighting, contamination and previously unseen defect types.
Evaluation
Missed defects by severity, false rejects, localization accuracy, inspection time and performance on held out parts and conditions.
Quality Prediction and Root Cause Analysis in Industrial Processes
Associate upstream process measurements, material identities, machine settings and tool histories with downstream quality. Investigate time lag alignment, confounding and process knowledge to rank plausible causes. Use controlled changes where feasible to distinguish causal evidence from correlation.
Research questions
Can warnings arrive before a defective batch is completed? Which suspected causes remain valid after production mix and equipment state change?
Proposed demonstration
Use a small production testbed or partner dataset, then evaluate a permitted process adjustment against a matched baseline.
Evaluation
Warning lead time, prediction calibration, validated causes, scrap and rework changes, and quality improvement after intervention.
Structural Health and Geotechnical Condition Intelligence
Combine strain, displacement, tilt, vibration, moisture and environmental measurements for bridges, foundations, retaining structures and industrial installations. Separate temperature and load effects from deterioration. Explore change detection and sparse sensing informed by structural models.
Research questions
How can baseline behavior adapt to seasons without masking progressive damage? Which combinations of sensors provide enough evidence to prioritize an engineering inspection?
Proposed demonstration
Use a structural test piece or annotated monitoring dataset with controlled load changes and simulated sensor loss.
Evaluation
Detection sensitivity, localization of abnormal response, false alerts under environmental change, uncertainty and agreement with independent inspection findings.
Risk Based Maintenance Planning and Intervention Optimization
Convert condition estimates into inspection and maintenance priorities using asset criticality, uncertainty, spare parts, crew availability and production constraints. Compare scheduling policies under uncertain degradation and repair duration. Keep economic recommendations traceable to explicit assumptions.
Research questions
When is another inspection more valuable than immediate replacement? How should a planner balance the cost of planned downtime against uncertain failure consequences?
Proposed demonstration
Replay maintenance histories or simulate a small asset fleet, comparing periodic scheduling, condition thresholds and optimized planning.
Evaluation
Unplanned downtime, maintenance cost, deferred high risk work, schedule feasibility and sensitivity to incorrect failure estimates.